Experiments Sections 9.1, 9.2, 9.3, 9.4, 9.5 Lecture 19 Robb T. Koether Hampden-Sydney College Wed, Feb 17, 2016 Robb T. Koether (Hampden-Sydney College) ExperimentsSections 9.1, 9.2, 9.3, 9.4, 9.5 Wed, Feb 17, 2016 1 / 18
ExperimentsSections 9.1, 9.2, 9.3, 9.4, 9.5
Lecture 19
Robb T. Koether
Hampden-Sydney College
Wed, Feb 17, 2016
Robb T. Koether (Hampden-Sydney College) ExperimentsSections 9.1, 9.2, 9.3, 9.4, 9.5 Wed, Feb 17, 2016 1 / 18
Outline
1 Experiments vs. Observation
2 Factors and Treatments
3 Randomized Experiments
4 Assignment
Robb T. Koether (Hampden-Sydney College) ExperimentsSections 9.1, 9.2, 9.3, 9.4, 9.5 Wed, Feb 17, 2016 2 / 18
Outline
1 Experiments vs. Observation
2 Factors and Treatments
3 Randomized Experiments
4 Assignment
Robb T. Koether (Hampden-Sydney College) ExperimentsSections 9.1, 9.2, 9.3, 9.4, 9.5 Wed, Feb 17, 2016 3 / 18
Experiments vs. Observation
DefinitionAn observational study observes individuals and measures variableswithout influencing the responses.
DefinitionAn experiment applies a treatment to the individuals and observes ormeasures variables to see the effect of the treatment.
In order to observe a cause-and-effect relationship, an experimentis much better than an observational study.Why?
Robb T. Koether (Hampden-Sydney College) ExperimentsSections 9.1, 9.2, 9.3, 9.4, 9.5 Wed, Feb 17, 2016 4 / 18
Experiments vs. Observation
DefinitionAn observational study observes individuals and measures variableswithout influencing the responses.
DefinitionAn experiment applies a treatment to the individuals and observes ormeasures variables to see the effect of the treatment.
In order to observe a cause-and-effect relationship, an experimentis much better than an observational study.
Why?
Robb T. Koether (Hampden-Sydney College) ExperimentsSections 9.1, 9.2, 9.3, 9.4, 9.5 Wed, Feb 17, 2016 4 / 18
Experiments vs. Observation
DefinitionAn observational study observes individuals and measures variableswithout influencing the responses.
DefinitionAn experiment applies a treatment to the individuals and observes ormeasures variables to see the effect of the treatment.
In order to observe a cause-and-effect relationship, an experimentis much better than an observational study.Why?
Robb T. Koether (Hampden-Sydney College) ExperimentsSections 9.1, 9.2, 9.3, 9.4, 9.5 Wed, Feb 17, 2016 4 / 18
Outline
1 Experiments vs. Observation
2 Factors and Treatments
3 Randomized Experiments
4 Assignment
Robb T. Koether (Hampden-Sydney College) ExperimentsSections 9.1, 9.2, 9.3, 9.4, 9.5 Wed, Feb 17, 2016 5 / 18
Example
Example (Treatments)In an attempt to determine which of two Statistics books helps thestudents learn better, an experiment was designed.One section of Statistics used the book “Statistics in Practice.”Another section used the book “Basic Statistics.”At the end of the semester, the average final grade of each sectionwas calculated.
What are the explanatory and response variables?Describe the “treatment” applied.
Robb T. Koether (Hampden-Sydney College) ExperimentsSections 9.1, 9.2, 9.3, 9.4, 9.5 Wed, Feb 17, 2016 6 / 18
Example
Example (Treatments)In an attempt to determine which of two Statistics books helps thestudents learn better, an experiment was designed.One section of Statistics used the book “Statistics in Practice.”Another section used the book “Basic Statistics.”At the end of the semester, the average final grade of each sectionwas calculated.What are the explanatory and response variables?
Describe the “treatment” applied.
Robb T. Koether (Hampden-Sydney College) ExperimentsSections 9.1, 9.2, 9.3, 9.4, 9.5 Wed, Feb 17, 2016 6 / 18
Example
Example (Treatments)In an attempt to determine which of two Statistics books helps thestudents learn better, an experiment was designed.One section of Statistics used the book “Statistics in Practice.”Another section used the book “Basic Statistics.”At the end of the semester, the average final grade of each sectionwas calculated.What are the explanatory and response variables?Describe the “treatment” applied.
Robb T. Koether (Hampden-Sydney College) ExperimentsSections 9.1, 9.2, 9.3, 9.4, 9.5 Wed, Feb 17, 2016 6 / 18
Confounded Variables
In the previous example, if a difference in final grades wasobserved, what might be the explanation?
The choice of textbook?The professor?The class size?The time when the class met?Which semester?Background of the students?
We say that the explanatory variable (the textbook) is confoundedby these other variables.
Robb T. Koether (Hampden-Sydney College) ExperimentsSections 9.1, 9.2, 9.3, 9.4, 9.5 Wed, Feb 17, 2016 7 / 18
Confounded Variables
In the previous example, if a difference in final grades wasobserved, what might be the explanation?
The choice of textbook?
The professor?The class size?The time when the class met?Which semester?Background of the students?
We say that the explanatory variable (the textbook) is confoundedby these other variables.
Robb T. Koether (Hampden-Sydney College) ExperimentsSections 9.1, 9.2, 9.3, 9.4, 9.5 Wed, Feb 17, 2016 7 / 18
Confounded Variables
In the previous example, if a difference in final grades wasobserved, what might be the explanation?
The choice of textbook?The professor?
The class size?The time when the class met?Which semester?Background of the students?
We say that the explanatory variable (the textbook) is confoundedby these other variables.
Robb T. Koether (Hampden-Sydney College) ExperimentsSections 9.1, 9.2, 9.3, 9.4, 9.5 Wed, Feb 17, 2016 7 / 18
Confounded Variables
In the previous example, if a difference in final grades wasobserved, what might be the explanation?
The choice of textbook?The professor?The class size?
The time when the class met?Which semester?Background of the students?
We say that the explanatory variable (the textbook) is confoundedby these other variables.
Robb T. Koether (Hampden-Sydney College) ExperimentsSections 9.1, 9.2, 9.3, 9.4, 9.5 Wed, Feb 17, 2016 7 / 18
Confounded Variables
In the previous example, if a difference in final grades wasobserved, what might be the explanation?
The choice of textbook?The professor?The class size?The time when the class met?
Which semester?Background of the students?
We say that the explanatory variable (the textbook) is confoundedby these other variables.
Robb T. Koether (Hampden-Sydney College) ExperimentsSections 9.1, 9.2, 9.3, 9.4, 9.5 Wed, Feb 17, 2016 7 / 18
Confounded Variables
In the previous example, if a difference in final grades wasobserved, what might be the explanation?
The choice of textbook?The professor?The class size?The time when the class met?Which semester?
Background of the students?
We say that the explanatory variable (the textbook) is confoundedby these other variables.
Robb T. Koether (Hampden-Sydney College) ExperimentsSections 9.1, 9.2, 9.3, 9.4, 9.5 Wed, Feb 17, 2016 7 / 18
Confounded Variables
In the previous example, if a difference in final grades wasobserved, what might be the explanation?
The choice of textbook?The professor?The class size?The time when the class met?Which semester?Background of the students?
We say that the explanatory variable (the textbook) is confoundedby these other variables.
Robb T. Koether (Hampden-Sydney College) ExperimentsSections 9.1, 9.2, 9.3, 9.4, 9.5 Wed, Feb 17, 2016 7 / 18
Confounded Variables
In the previous example, if a difference in final grades wasobserved, what might be the explanation?
The choice of textbook?The professor?The class size?The time when the class met?Which semester?Background of the students?
We say that the explanatory variable (the textbook) is confoundedby these other variables.
Robb T. Koether (Hampden-Sydney College) ExperimentsSections 9.1, 9.2, 9.3, 9.4, 9.5 Wed, Feb 17, 2016 7 / 18
Experiments
Definition (Subjects)In an experiment, the individuals are called subjects.
Definition (Factors)In an experiment, the explanatory variables are called factors.
Definition (Treatment)In an experiment, a treatment is a specific combination of values of thefactors.
Robb T. Koether (Hampden-Sydney College) ExperimentsSections 9.1, 9.2, 9.3, 9.4, 9.5 Wed, Feb 17, 2016 8 / 18
Example
Example (Treatments)Suppose we performed the textbook experiment with twoprofessors: Prof. Smith and Prof. Jones.If we used two sections and
Prof. Smith used “Statistics in Practice”Prof. Jones use “Basic Statistics”
How many treatments are there?
Would there be a confounding of variables?
Robb T. Koether (Hampden-Sydney College) ExperimentsSections 9.1, 9.2, 9.3, 9.4, 9.5 Wed, Feb 17, 2016 9 / 18
Example
Example (Treatments)Suppose we performed the textbook experiment with twoprofessors: Prof. Smith and Prof. Jones.If we used two sections and
Prof. Smith used “Statistics in Practice”Prof. Jones use “Basic Statistics”
How many treatments are there?Would there be a confounding of variables?
Robb T. Koether (Hampden-Sydney College) ExperimentsSections 9.1, 9.2, 9.3, 9.4, 9.5 Wed, Feb 17, 2016 9 / 18
Example
Example (Treatments)If we used four sections and
Each professor taught two sectionsEach professor used one book in one section and the other book inthe other section.
How many treatments are there?
Would there be a confounding of the two explanatory variables?
Robb T. Koether (Hampden-Sydney College) ExperimentsSections 9.1, 9.2, 9.3, 9.4, 9.5 Wed, Feb 17, 2016 10 / 18
Example
Example (Treatments)If we used four sections and
Each professor taught two sectionsEach professor used one book in one section and the other book inthe other section.
How many treatments are there?Would there be a confounding of the two explanatory variables?
Robb T. Koether (Hampden-Sydney College) ExperimentsSections 9.1, 9.2, 9.3, 9.4, 9.5 Wed, Feb 17, 2016 10 / 18
Example
Example (Treatments)If we used eight sections and
Each professor taught four sectionsEach professor used one book in two sections and the other bookin the other two sections.
How many treatments are there?
Robb T. Koether (Hampden-Sydney College) ExperimentsSections 9.1, 9.2, 9.3, 9.4, 9.5 Wed, Feb 17, 2016 11 / 18
Outline
1 Experiments vs. Observation
2 Factors and Treatments
3 Randomized Experiments
4 Assignment
Robb T. Koether (Hampden-Sydney College) ExperimentsSections 9.1, 9.2, 9.3, 9.4, 9.5 Wed, Feb 17, 2016 12 / 18
Definition (Randomized Experiment)A randomized experiment is one in which the subjects are assigned atrandom to the different groups.
Why use a randomized experiment?
Robb T. Koether (Hampden-Sydney College) ExperimentsSections 9.1, 9.2, 9.3, 9.4, 9.5 Wed, Feb 17, 2016 13 / 18
Definition (Randomized Experiment)A randomized experiment is one in which the subjects are assigned atrandom to the different groups.
Why use a randomized experiment?
Robb T. Koether (Hampden-Sydney College) ExperimentsSections 9.1, 9.2, 9.3, 9.4, 9.5 Wed, Feb 17, 2016 13 / 18
Example
Example (Randomized Experiment)Suppose we used four sections and the design
Prof.Smith
Statisticsin Practice
BasicStatistics
Prof.Jones
25 25
25 25
We could randomly assign a total of 100 students to the fourtreatments.Describe exactly how we would do that.
Robb T. Koether (Hampden-Sydney College) ExperimentsSections 9.1, 9.2, 9.3, 9.4, 9.5 Wed, Feb 17, 2016 14 / 18
Example
Example (Randomized Experiment)There is more than one way.
One possibility:
Number the students 1 to 100.Use randInt(1,100) fifty times to assign 50 students to Prof.Smith.The rest go to Prof. Jones.Renumber each group 1 to 50.Use randInt(1,50) twenty-five times to assign 25 students ineach group to “Statistics in Practice.”The rest use “Basic Statistics.”
Robb T. Koether (Hampden-Sydney College) ExperimentsSections 9.1, 9.2, 9.3, 9.4, 9.5 Wed, Feb 17, 2016 15 / 18
Example
Example (Randomized Experiment)There is more than one way.One possibility:
Number the students 1 to 100.Use randInt(1,100) fifty times to assign 50 students to Prof.Smith.The rest go to Prof. Jones.Renumber each group 1 to 50.Use randInt(1,50) twenty-five times to assign 25 students ineach group to “Statistics in Practice.”The rest use “Basic Statistics.”
Robb T. Koether (Hampden-Sydney College) ExperimentsSections 9.1, 9.2, 9.3, 9.4, 9.5 Wed, Feb 17, 2016 15 / 18
Example
Example (Randomized Experiment)There is more than one way.One possibility:
Number the students 1 to 100.
Use randInt(1,100) fifty times to assign 50 students to Prof.Smith.The rest go to Prof. Jones.Renumber each group 1 to 50.Use randInt(1,50) twenty-five times to assign 25 students ineach group to “Statistics in Practice.”The rest use “Basic Statistics.”
Robb T. Koether (Hampden-Sydney College) ExperimentsSections 9.1, 9.2, 9.3, 9.4, 9.5 Wed, Feb 17, 2016 15 / 18
Example
Example (Randomized Experiment)There is more than one way.One possibility:
Number the students 1 to 100.Use randInt(1,100) fifty times to assign 50 students to Prof.Smith.
The rest go to Prof. Jones.Renumber each group 1 to 50.Use randInt(1,50) twenty-five times to assign 25 students ineach group to “Statistics in Practice.”The rest use “Basic Statistics.”
Robb T. Koether (Hampden-Sydney College) ExperimentsSections 9.1, 9.2, 9.3, 9.4, 9.5 Wed, Feb 17, 2016 15 / 18
Example
Example (Randomized Experiment)There is more than one way.One possibility:
Number the students 1 to 100.Use randInt(1,100) fifty times to assign 50 students to Prof.Smith.The rest go to Prof. Jones.
Renumber each group 1 to 50.Use randInt(1,50) twenty-five times to assign 25 students ineach group to “Statistics in Practice.”The rest use “Basic Statistics.”
Robb T. Koether (Hampden-Sydney College) ExperimentsSections 9.1, 9.2, 9.3, 9.4, 9.5 Wed, Feb 17, 2016 15 / 18
Example
Example (Randomized Experiment)There is more than one way.One possibility:
Number the students 1 to 100.Use randInt(1,100) fifty times to assign 50 students to Prof.Smith.The rest go to Prof. Jones.Renumber each group 1 to 50.
Use randInt(1,50) twenty-five times to assign 25 students ineach group to “Statistics in Practice.”The rest use “Basic Statistics.”
Robb T. Koether (Hampden-Sydney College) ExperimentsSections 9.1, 9.2, 9.3, 9.4, 9.5 Wed, Feb 17, 2016 15 / 18
Example
Example (Randomized Experiment)There is more than one way.One possibility:
Number the students 1 to 100.Use randInt(1,100) fifty times to assign 50 students to Prof.Smith.The rest go to Prof. Jones.Renumber each group 1 to 50.Use randInt(1,50) twenty-five times to assign 25 students ineach group to “Statistics in Practice.”
The rest use “Basic Statistics.”
Robb T. Koether (Hampden-Sydney College) ExperimentsSections 9.1, 9.2, 9.3, 9.4, 9.5 Wed, Feb 17, 2016 15 / 18
Example
Example (Randomized Experiment)There is more than one way.One possibility:
Number the students 1 to 100.Use randInt(1,100) fifty times to assign 50 students to Prof.Smith.The rest go to Prof. Jones.Renumber each group 1 to 50.Use randInt(1,50) twenty-five times to assign 25 students ineach group to “Statistics in Practice.”The rest use “Basic Statistics.”
Robb T. Koether (Hampden-Sydney College) ExperimentsSections 9.1, 9.2, 9.3, 9.4, 9.5 Wed, Feb 17, 2016 15 / 18
Randomized Experiments
Would the randomized design just described eliminate theconfounding of variables?
Why not?What further steps could we take?
Robb T. Koether (Hampden-Sydney College) ExperimentsSections 9.1, 9.2, 9.3, 9.4, 9.5 Wed, Feb 17, 2016 16 / 18
Randomized Experiments
Would the randomized design just described eliminate theconfounding of variables?Why not?
What further steps could we take?
Robb T. Koether (Hampden-Sydney College) ExperimentsSections 9.1, 9.2, 9.3, 9.4, 9.5 Wed, Feb 17, 2016 16 / 18
Randomized Experiments
Would the randomized design just described eliminate theconfounding of variables?Why not?What further steps could we take?
Robb T. Koether (Hampden-Sydney College) ExperimentsSections 9.1, 9.2, 9.3, 9.4, 9.5 Wed, Feb 17, 2016 16 / 18
Outline
1 Experiments vs. Observation
2 Factors and Treatments
3 Randomized Experiments
4 Assignment
Robb T. Koether (Hampden-Sydney College) ExperimentsSections 9.1, 9.2, 9.3, 9.4, 9.5 Wed, Feb 17, 2016 17 / 18
Assignment
AssignmentRead Sections 9.1, 9.2, 9.3, 9.4.Apply Your Knowledge: 1, 2, 4, 5, 10.Check Your Skills: 19, 20, 21, 22.Exercises 29, 31, 32, 33, 34.
Robb T. Koether (Hampden-Sydney College) ExperimentsSections 9.1, 9.2, 9.3, 9.4, 9.5 Wed, Feb 17, 2016 18 / 18